Fairness in Incomplete Information Bargaining: Theory and Widespread Evidence from the Field
This paper uses detailed data on sequential offers from seven vastly different real-world bargaining settings to document a robust pattern: agents favor offers that split the difference between the two most recent offers on the table. Our settings include negotiations for used cars, insurance injury claims, a TV game show, auto rickshaw rides, housing, international trade tariffs, and online retail. We demonstrate that this pattern can arise in a perfect Bayesian equilibrium of an alternating-offer game with two-sided incomplete information, but this equilibrium is far from unique. We then provide a robust-inference argument to explain why agents may view the two most recent offers as corresponding to the potential surplus. Split-the-difference offers under this weaker, robust inference can then be viewed as fair. We present a number of other patterns in each data setting that point to split-the-difference offers as a strong social norm, whether in high-stakes or low-stakes negotiations.